Trauma remains a leading cause of morbidity and mortality across all age groups and is the most frequent cause of death and disability in individuals under eighteen years (1). Although the passage of time is often considered beneficial for post-injury recovery, many traumatic injuries do not improve without timely assessment and management (2). Consequently, multiple trauma-scoring systems have been developed and are widely applied in trauma settings, including the Glasgow Coma Scale (GCS), Acute Physiology and Chronic Health Evaluation II (APACHE-II), Revised Trauma Score (RTS), Injury Severity Score (ISS), and Trauma and Injury Severity Score (TRISS) (3). These tools support pre-hospital assessment, in-hospital evaluation, prognostication, and decision-making related to patient triage and transport (4).

Among these, ISS is one of the most frequently utilized trauma-scoring systems (5). summarizing overall injury severity by integrating the three most severely affected body regions (6). However, due to anatomical and physiological differences between children and adults, ISS may overestimate mortality risk in pediatric populations (7). Contributing to ongoing debate regarding its suitability in children Increasingly, RTS is used in early triage and severity determination (8). while TRISS combines RTS, ISS, and age-related factors to predict survival (9). and remains one of the most validated prognostic tools despite computational limitations (10). Other composite scores, such as MGAP, further support the relevance of physiological indicators and age in outcome prediction (11).

Currently, no universally validated pediatric-specific trauma-severity score exists (12). and ISS remains most commonly applied despite heterogeneous injury mechanisms in children  (13). Therefore, this study compares the predictive accuracy of ISS, RTS, and TRISS for mortality and morbidity in pediatric trauma, aiming to identify superior prognostic tools to support early risk stratification, clinical decision-making and improved outcomes in this vulnerable population.

This was an observational study conducted in the Pediatric Emergency of Dr. Ziauddin University Hospital, North Campus and involved pediatric patients with a history of trauma within a period of 12 months. The target population was both male and female children aged between 1-15 years that visited the emergency department within 24 hours of incurring trauma. The convenience sampling method was adopted. The patients who received management elsewhere before, or failure to consent by parents or guardians were not included in the study. The study protocol was approved by IRB of hospital via reference number A structured proforma was used to gather data which included demographic data, presenting complaints, mechanism of injury, vital signs, oxygen saturation and neurological status upon presentation. The level of trauma was measured by using well-known scoring systems, such as ISS, RTS, and TRISS. The entirety was obtained during the first evaluation during emergencies, so that there could be consistency and reduces bias. The main outcomes evaluated included mortality and morbidity where morbidity was taken as composite to mean the incidence of severe clinical outcomes like the necessity to be admitted to intensive care require surgery or stay in the hospital after a long period. SPSS v 27 was used for data analysis. Patient characteristics and trauma scores were summarized with the help of descriptive statistics. The chi-square test was used to first determine any association between trauma scores and outcome. A binary logistic regression analysis was subsequently used to estimate the independent predictive mortality and morbidity of each of the trauma scoring systems. Receiver operating characteristic (ROC) curve analysis was conducted to assess the diagnostic performance of each score for mortality and morbidity. All the analyses were deemed to be statistically significant at a p-value of 0.05 or below.

The mean age of the cohort was 8.08 ± 4.87 years, with non-survivors being significantly younger than survivors (3.45 ± 3.78 vs. 8.51 ± 4.86 years; p < 0.001). A statistically significant difference in gender distribution was observed, with males comprising a higher proportion of survivors, while females constituted the majority of non-survivors (p < 0.001).

No significant differences were noted between survivors and non-survivors with respect to the nature or cause of trauma (p = 0.525 for both). In contrast, marked and statistically significant differences were observed across key physiological parameters at presentation. Non-survivors demonstrated significantly lower oxygen saturation and blood pressure values, alongside higher heart rates (p < 0.001 for all). Respiratory rate was also significantly reduced among non-survivors (p < 0.001).

Injury severity measures differed substantially between groups. Non-survivors exhibited significantly lower GCS and Revised

Trauma Score values, as well as higher Injury Severity Scores, compared with survivors (all p < 0.001). Similarly, Trauma and Injury Severity Score values were significantly lower among non-survivors (p < 0.001). Length of hospital stay did not differ significantly between groups (p = 0.078). The proportion of patients experiencing morbidity did not differ significantly between survivors and non-survivors (p = 0.590). Table 1 presents the details, stratified by survival status.

Table 1:  Baseline Clinical and Injury Characteristics of Patients Stratified by Mortality Outcome

NC, not computable

The optimal cut-off values and corresponding diagnostic performance for mortality prediction were: GCS ≤ 10 (sensitivity 53.8%, specificity 94.7%), ISS ≥ 16 (sensitivity 38.5%, specificity 72.0%), RTS ≤ 5.5 (sensitivity 69.2%, specificity 99.2%), and TRISS ≤ 96% (sensitivity 61.5%, specificity 95.5%).

In Table 2 multivariable logistic regression analysis shows that the age and male gender were significantly associated with mortality in the GCS, ISS, and TRISS models (p < 0.05), but not in the RTS model. Lower GCS was independently associated with increased mortality (aOR = 0.69, 95% CI: 0.56–0.85; p < 0.001). Higher ISS showed a significant increase in mortality risk (aOR = 1.08, 95% CI: 1.03–1.13; p = 0.002). Lower RTS was strongly associated with mortality (aOR = 0.12, 95% CI: 0.04–0.38; p < 0.001). TRISS was also a significant predictor, with decreasing scores associated with higher mortality (aOR = 0.91, 95% CI: 0.83–0.98; p = 0.017).

Table 2: Multivariable Logistic Regression Analysis of Mortality Predictors across Trauma Score Models

In Table 3 shows that patients with morbidity showed a significantly higher proportion of males compared with those without morbidity (86.4% vs. 62.1%, p = 0.003). The distribution of nature and cause of trauma differed significantly between groups (p = 0.026), with extremity and soft tissue injuries more frequent among patients with morbidity. Injury Severity Score was significantly higher in the morbidity group (14.91 ± 8.15 vs. 7.00 ± 2.80; p < 0.001), as was the length of hospital stay (24.12 ± 38.03 vs. 8.02 ± 13.83; p = 0.030). Follow-up requirement was markedly higher among patients with morbidity (97.1% vs. 37.9%; p < 0.001).

Table 3: Comparison of Clinical and Injury Characteristics by Morbidity Status

A GCS cut-off of ≤ 14 demonstrated high sensitivity (81.6%) with perfect specificity (100%). ISS showed the highest sensitivity at a threshold of ≥ 7 (90.3%), though with lower specificity (44.8%). RTS, at a cut-off of ≤ 7.5, yielded high sensitivity (89.3%) but very low specificity (3.4%). TRISS exhibited a balanced predictive performance, with a cut-off of ≤ 98.5% achieving high sensitivity (88.3%) and specificity (96.6%).

Table 4 presents predictors of morbidity across GCS-, ISS-, RTS-, and TRISS-based models. Age was not significantly associated with morbidity in any model (all p > 0.05). Male gender showed a consistent and statistically significant association with higher odds of morbidity across all models, with aORs ranging from 3.31 to 4.27 (p ≤ 0.044). Within score-specific predictors, ISS was independently associated with increased morbidity in the ISS model (aOR 1.38, 95% CI: 1.19–1.60; p < 0.001). RTS did not demonstrate a significant association with morbidity (p = 0.158). In the TRISS model, lower TRISS values were significantly associated with morbidity (aOR 0.12, 95% CI: 0.02–0.81; p = 0.029). GCS was not a significant predictor in the GCS-based model (p = 0.995).

Table 4:  Multivariable Logistic Regression Analysis for Predictors of Morbidity Using Different Trauma Score Models

The figure 1 displays ROC curves comparing the discriminatory performance of GCS, ISS, RTS, and TRISS for predicting mortality (left panel) and morbidity (right panel). For mortality, RTS and ISS demonstrate comparatively better discrimination, whereas GCS and TRISS show limited predictive ability. For morbidity, ISS and TRISS exhibit superior performance above the reference line, while GCS shows modest discrimination and RTS demonstrates poor specificity across thresholds.

Figure 1:  ROC Curves for Trauma Scores Predicting Mortality and Morbidity

The study determined the predictive validity of four systems of trauma-scoring including GCS, ISS, RTS, and TRISS against mortality and morbidity. There were major differences in the number of survivors and non-survivors in terms of demographic and physiological variables. Non-survivors were also younger and showed less oxygen saturation, decreased blood pressure, increased heart rate, and low GCS and RTS scores, as well as, increased ISS and decreased TRISS values. The receiver operating characteristic revealed that TRISS and ISS were useful to predict mortality with the best cut-off points. Advanced age, male gender, diminished GCS, increased ISS, decreased RTS and TRISS became significant predictors of mortality. Besides, sex and increased ISS were important morbidity predictors, which support the greater prognosis of combined physiological and anatomical scoring systems, especially TRISS and ISS, in comparison with physiological indices alone in pediatric trauma. The latest data is always in favor of the efficiency of combined trauma scoring systems in outcome prediction in pediatric groups. Dundar et al. (14) have shown that the Injury Severity Score (ISS) had better discriminatory potential to mortality, and the Revised Trauma Score (RTS) had independent outcome risk-predictive value and was better than isolated physiological variables. Multicenter research by Khavandegar et al. (15) also revealed that TRISS and adjusted TRISS (aTRISS) showed that TRISS and adjusted TRISS (aTRISS) produced the highest area-under-curve (AUC) which was significantly higher than individual scores including GCS, RTS and ISS. These results support the current study, in which TRISS became superior to physiological scoring, only, and emphasized the importance of using anatomical and physiological measures. Similarly, Toida et al. (16) who used a national sample of pediatric blunt trauma patients, validated the outstanding discriminative ability of TRISS (AUC ≥0.97) among age groups, even compared to other emerging predictive models. All these studies support the use of anatomical -physiological composite scoring systems as stronger prognostic variables in pediatric trauma which agrees with the finding that TRISS is a very effective outcome stratification index. Similar findings were reported in other studies on pediatric trauma, where the authors including Çelegen et al. (17) verified that ISS is more effective than other systems including PTS, BIG score, and PRISM-III, and that the ISS cutoff of 20.5 reported a high sensitivity in predicting mortality. Dundar et al. (14) have also found that both ISS and RTS are independent predictors of mortality and have high values of AUC (0.98 and 0.83 respectively) and can be used as prognostic factors in multivariate analyses. Khavandegar et al. (15) also affirmed that TRISS and aTRISS proved to be good predictors of mortality but established ISS as a good tool when compared to other scores like NISS and AIS. All these results support the current study by establishing that the use of anatomical and physiological parameters like ISS with RTS or TRISS has a significant role in predicting mortality and stratifying outcomes in pediatric trauma. This study showed that trauma severity was a significant predictor of ICU admission and complications in regard to morbidity, similar to recent pediatric evidence. The predictive ability of both NTRISS and TRISS was found to be high in predicting ICU admission, a key indicative of morbidity, at 0.837 and 0.823 respectively, which is higher than the physiological scores (15). Also, pediatric ICU trauma admission studies have identified that increased ISS is associated with mortality, mechanical ventilation, and longer ICU, which supports the severity of injuries and poor outcomes (18). On the same note, ISS greater than 15, lower GCS, and multisystem were also predictors of ICU admission in pediatric blunt abdominal trauma (19). All these results emphasize the relevance of the scores of anatomical severity, especially ISS and TRISS to the risk of morbidity because the higher the ISS and the lower the TRISS, the higher the morbidity. The combined scoring methods are also supported by the findings in adult and mixed populations of trauma. Murtaza Kaya et al. (20) showed that TRISS and GCS predict mortality effectively in adults with trauma as compared to single indices. Further comparative diagnostic studies have also shown a superiority of TRISS to RTISS, GCS and ISS in ROC performance (21). and systematic studies also supported TRISS as a highly predictive model using a combination of physiological, anatomical and age factors (22). Even though the physiology of pediatric trauma is not similar to that of adults, such tendencies support the importance of the universal scoring system in all age groups. System limitations are identified in a few studies even though there is good evidence. Pediatric trauma scoring scales include PTS and GCS which might need recalibration across groups and ideal cut-offs differ by geography and injury profile (23). In line with this, some cohort-specific thresholds including ISS 16 and TRISS 96 were found in this study, which highlights the situational-specificity of predictive scoring. Although ROC analysis is still a gold standard in assessing the discriminating power, the variability in the event rates especially in low-mortality pediatric cohort requires additional caution in cut-off values according to clinical and demographic situation. Limitation The limitations of our research include a rather low amount of mortality events, which restrains the accuracy of mortality prediction models and multivariable analyses. Also, the limited sample size of the research (single center) can limit the representatively of our results. Future studies should thus engage in multicenter studies with more significant pediatric groups and possibly even multicenter studies that incorporate real-time predictive models that incorporate machine learning methods because current evidence is suggesting that these methods could be superior to conventional scoring mechanisms.

The study determined the predictive validity of four systems of trauma-scoring including GCS, ISS, RTS, and TRISS against mortality and morbidity. There were major differences in the number of survivors and non-survivors in terms of demographic and physiological variables. Non-survivors were also younger and showed less oxygen saturation, decreased blood pressure, increased heart rate, and low GCS and RTS scores, as well as, increased ISS and decreased TRISS values. The receiver operating characteristic revealed that TRISS and ISS were useful to predict mortality with the best cut-off points. Advanced age, male gender, diminished GCS, increased ISS, decreased RTS and TRISS became significant predictors of mortality. Besides, sex and increased ISS were important morbidity predictors, which support the greater prognosis of combined physiological and anatomical scoring systems, especially TRISS and ISS, in comparison with physiological indices alone in pediatric trauma.

The latest data is always in favor of the efficiency of combined trauma scoring systems in outcome prediction in pediatric groups. Dundar et al. (14) have shown that the Injury Severity Score (ISS) had better discriminatory potential to mortality, and the Revised Trauma Score (RTS) had independent outcome risk-predictive value and was better than isolated physiological variables. Multicenter research by Khavandegar et al. (15)  also revealed that TRISS and adjusted TRISS (aTRISS) showed that TRISS and adjusted TRISS (aTRISS) produced the highest area-under-curve (AUC) which was significantly higher than individual scores including GCS, RTS and ISS. These results support the current study, in which TRISS became superior to physiological scoring, only, and emphasized the importance of using anatomical and physiological measures. Similarly, Toida et al. (16) who used a national sample of pediatric blunt trauma patients, validated the outstanding discriminative ability of TRISS (AUC ≥0.97) among age groups, even compared to other emerging predictive models. All these studies support the use of anatomical -physiological composite scoring systems as stronger prognostic variables in pediatric trauma which agrees with the finding that TRISS is a very effective outcome stratification index.

Similar findings were reported in other studies on pediatric trauma, where the authors including Çelegen et al. (17) verified that ISS is more effective than other systems including PTS, BIG score, and PRISM-III, and that the ISS cutoff of 20.5 reported a high sensitivity in predicting mortality. Dundar et al.  (14) have also found that both ISS and RTS are independent predictors of mortality and have high values of AUC (0.98 and 0.83 respectively) and can be used as prognostic factors in multivariate analyses. Khavandegar et al. (15) also affirmed that TRISS and aTRISS proved to be good predictors of mortality but established ISS as a good tool when compared to other scores like NISS and AIS. All these results support the current study by establishing that the use of anatomical and physiological parameters like ISS with RTS or TRISS has a significant role in predicting mortality and stratifying outcomes in pediatric trauma.

This study showed that trauma severity was a significant predictor of ICU admission and complications in regard to morbidity, similar to recent pediatric evidence. The predictive ability of both NTRISS and TRISS was found to be high in predicting ICU admission, a key indicative of morbidity, at 0.837 and 0.823 respectively, which is higher than the physiological scores (15). Also, pediatric ICU trauma admission studies have identified that increased ISS is associated with mortality, mechanical ventilation, and longer ICU, which supports the severity of injuries and poor outcomes (18). On the same note, ISS greater than 15, lower GCS, and multisystem were also predictors of ICU admission in pediatric blunt abdominal trauma (19). All these results emphasize the relevance of the scores of anatomical severity, especially ISS and TRISS to the risk of morbidity because the higher the ISS and the lower the TRISS, the higher the morbidity.

The combined scoring methods are also supported by the findings in adult and mixed populations of trauma. Murtaza Kaya et al. (20)  showed that TRISS and GCS predict mortality effectively in adults with trauma as compared to single indices. Further comparative diagnostic studies have also shown a superiority of TRISS to RTISS, GCS and ISS in ROC performance (21). and systematic studies also supported TRISS as a highly predictive model using a combination of physiological, anatomical and age factors (22). Even though the physiology of pediatric trauma is not similar to that of adults, such tendencies support the importance of the universal scoring system in all age groups.

System limitations are identified in a few studies even though there is good evidence. Pediatric trauma scoring scales include PTS and GCS which might need recalibration across groups and ideal cut-offs differ by geography and injury profile (23). In line with this, some cohort-specific thresholds including ISS 16 and TRISS 96 were found in this study, which highlights the situational-specificity of predictive scoring. Although ROC analysis is still a gold standard in assessing the discriminating power, the variability in the event rates especially in low-mortality pediatric cohort requires additional caution in cut-off values according to clinical and demographic situation.

Limitation

The limitations of our research include a rather low amount of mortality events, which restrains the accuracy of mortality prediction models and multivariable analyses. Also, the limited sample size of the research (single center) can limit the representatively of our results. Future studies should thus engage in multicenter studies with more significant pediatric groups and possibly even multicenter studies that incorporate real-time predictive models that incorporate machine learning methods because current evidence is suggesting that these methods could be superior to conventional scoring mechanisms.